15:59:28 #startmeeting kolla 15:59:29 Meeting started Wed Jun 28 15:59:28 2017 UTC and is due to finish in 60 minutes. The chair is inc0. Information about MeetBot at http://wiki.debian.org/MeetBot. 15:59:31 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 15:59:34 The meeting name has been set to 'kolla' 15:59:42 #topic w00t for Kolla! 15:59:56 few minutes to show the love guys;) 16:00:17 woot 16:00:21 o/ 16:00:26 w00t 16:00:27 woot 16:00:30 woot! 16:00:49 resistance is futile?:) 16:00:50 hello guys o/ w00t! 16:00:53 o/ 16:00:59 hey coolsvap how u been 16:01:03 welcome back coolsvap:) 16:01:14 hi coolsvap 16:01:22 sup coolsvap 16:01:25 woot 16:01:28 w00t 16:02:08 i am all well its good to be back :) 16:02:20 woot 16:02:22 https://vignette2.wikia.nocookie.net/uncyclopedia/images/a/ad/Drop_bear.JPG/revision/latest?cb=20050510111310&format=original 16:02:25 oh, hi coolsvap! 16:02:36 coolsvap is back! 16:02:56 we're all very happy to see you Swapnil 16:03:25 thanks inc0 :) 16:03:34 ok, let's get to meeting 16:03:35 sdake: +1 16:03:40 #topic Announcements 16:04:02 so happy day for our community, not only coolsvap is back, we also have new core member! 16:04:15 congratulations to spsurya_ and great job! 16:04:23 Congrats spsurya_ !! 16:04:27 spsurya_: (y) 16:04:52 congrats spsurya_ 16:04:58 congrats spsurya_ 16:05:06 any other announcements? 16:05:26 guess not:) 16:05:31 let's get to agenda 16:05:32 thanks all for votes and wishes 16:05:47 #topic Triaging of bugs for this week (spsurya) 16:06:01 I took liberty of removing k8s tool from this topic as we have dedicated one next 16:06:05 https://etherpad.openstack.org/p/kolla-bug-triage-week 16:06:15 #link https://etherpad.openstack.org/p/kolla-bug-triage-week 16:06:17 this got few one which are ready to fix 16:06:35 if anyone like they can pick from this 16:06:44 to conitrubte 16:07:39 What does status 'ready to fix' mean exactly? It's a triaged as a valid bug and looking for an owner? 16:07:42 more interesting one 16:07:51 tested rwellum tool 16:07:57 rwellum: yeah 16:08:13 rwellum: yes looking for owner 16:08:17 spsurya_: hold that thought, we have rwellum tool topic after this one 16:08:17 thanks 16:08:30 you can grab any if you like 16:08:38 to contribute more 16:08:50 also if anyone ever asks how to contribute 16:08:52 so on rwellum tool 16:08:56 this is great list to get people started 16:09:03 * krtaylor looks 16:09:04 right, I'll just start the topic;) 16:09:15 #topic Question - upstream a tool (or not)? rwellum 16:09:23 Intially found some imrovement in tools and communicated the same with rwellum. 16:09:23 thanks rewellum for keeping fix the monir issues. 16:09:23 rwellum spsurya_ go ahead 16:09:42 https://github.com/jascott1/arboreal/ I 16:09:44 Successfully tested AIO on centos with 2 NICs, 16GB RAM, 50GB HDD, 4CPUs resources. 16:09:48 I'll add this one too 16:10:00 rwellum: mind pasting link to your rtepo please? 16:10:18 inc0: IMO would be good if he push it to upstream and get some feedback more people 16:10:23 yes was good to get another (core) use it - before sending to everyone. 16:10:24 https://github.com/RichWellum/k8s 16:10:54 Failed with 1CPUS as well as with 2 CPUs that should not happen. 16:10:54 This may be the issue in kubernetes 1.6.5 16:11:10 as per rwellum IIRC 16:11:20 so let's discuss what we want to do in general 16:11:26 with this problem 16:11:54 inc0: Need to test on Ubuntu 16.04 too which is left AIO 16:11:58 we have Arboreal that sets up k8s with terraform+ansible 16:12:20 we also have playbook for kolla there 16:12:32 we have rwellum's tool 16:12:33 the terraform bits need a little more work but I am doing that now 16:12:46 imean they work but arent as portable as could be 16:12:49 apperantly there is need for that stuff:) 16:12:58 jascott1: 16:13:08 rwellum: tool also need some work 16:13:19 to get into final shape 16:14:24 i'd recommend focusing on one deployment mechanism rather than 3 parallel development paths :) 16:14:34 that too 16:14:54 i'm personally a fan of the microchart ansible approach - very flexible and complete control 16:14:55 but since we have at least 2 different 16:14:58 problem is real 16:15:05 Yeah that's why I raised it in the first place as an agenda item. 16:15:28 Don't want to deflect potential users and contributors from jascott1 excellent solution. 16:15:41 sdake: +1, yes we need one deployment mechanism 16:15:53 I was mainly through developing this tool when I saw jascott1 was doing this. 16:16:01 simple is good - i believe brad's approach is pretty solid - haven't looked at your solution jascott1 16:16:12 its the dev guide in ansible format 16:16:19 rwellum: neither are excellent, but I agree with sdake, we need good and easy method to move from clean k8s to kolla 16:17:05 jascott1 there were some earlly reviews on brad's approach - and there was some desire to keep things "seprated" 16:17:39 sdake: separated will be super easy when we have roles figured out 16:17:55 I haven't seen 'Brads' tool sdake - I know there is also a shell-script - I thought that was the second tool. 16:18:02 what i'm saying in essence is jascott1 bradjones can you guys team up plz :) 16:18:03 #link https://review.openstack.org/#/c/473588/8 16:18:10 at least between the two approaches we should have all/most of the functionality we need and be able to reference each 16:18:25 I'll also add this 16:18:29 #link https://review.openstack.org/#/c/474770/ 16:18:38 at some point we'll run Brad's playbook from inside container 16:18:44 and this might be the container 16:19:37 we can create helm chart to run playbooks:) 16:19:55 helm install orchestration/generate-config 16:20:02 So these ansible approaches seem almost like they are heading towards production level usage? My tool is a simple single user, AIO on a VM. Good for learning k8s and kolla etc. 16:20:02 helm install orchestration/deploy-kolla 16:20:04 i like it -based upon kfox1111 's review of brad's work - the config gen should be a seprate thing i htink 16:20:19 rwellum yes - these are production deployment tools 16:20:28 (or will be ;) 16:20:33 I kinda disagree, it has to be optional 16:20:42 but in kolla-ansible config gen is part of deployment 16:20:44 and I like that 16:20:52 if you want to provide your own configs, fine 16:21:15 here is what i'd recommend with the approach to solving this problem 16:21:27 we know for certain deployment is one type of "playbook" needed 16:21:35 lets get that working well in a docker container launched by a helm install 16:21:46 and then sort out the next part fo the solution next 16:21:49 that's what I propose... 16:21:57 and that's what this change is about.. 16:22:00 if we try to solve too many porblems at once, we end with sea salt production 16:23:27 what I'm trying to say 16:23:36 is when we have docker image with ansible in it 16:23:39 we can do all 16:23:49 right makes sense 16:25:14 kfox1111 around? 16:26:17 ok, I guess we need to get kfox1111 and Serguey on board 16:26:33 let's talk to them later today about that 16:26:53 in the meantime, I'll work on dockerfile part + configmap generation, that's coming on nicely 16:27:04 and Brad's playbook should be easy afterwards 16:27:27 to be put into container that is 16:28:09 anything else on the topic? 16:28:38 guess not 16:28:45 #topic Open discussion 16:28:52 one thing from me 16:28:55 I guess with both ansible solutions floating around, the general consensus is my tool is probably not upstream-worthy at this point. So occasionally we have new folks on IRC struggling to bring up kolla-k8s. Like 'ddejog'. I should point him to jascott1 solution then? 16:29:15 Sorry inc0 - slow typer here 16:29:22 rwellum you can push a review - and then people can use it 16:29:24 rwellum: feel free to suggest your own tool, no worries about that 16:29:39 Oh ok then. 16:29:42 what we're sayin is it's a prothesis and we need something better which is in the making by Brad and myself 16:29:44 i think we have decided to boil this sepcific ocean with ansible however :) 16:29:48 Cheers guys. 16:29:54 Yes agree fully with that. 16:30:10 so let's focus on proper way to solve it is what we're saying;) 16:30:46 from me, I need these changes merged for docker registry people;) https://review.openstack.org/#/c/475940/ https://review.openstack.org/#/c/475940/ 16:30:59 https://review.openstack.org/#/c/475469/ 16:31:33 infra change is blocked on these 2 and whole dockerhub publisher is blocked on this 16:31:43 so can I have some reviews please?:) 16:32:02 Two of those links are the same inc0 ? 16:32:19 https://review.openstack.org/#/c/475940/ https://review.openstack.org/#/c/475469/ 16:32:21 these two 16:33:01 ok 16:33:33 Will take a look. 16:34:00 thank you 16:34:08 anything else wants to say anything? 16:34:12 anyone* 16:34:15 o/ 16:34:20 go ahead Duong 16:34:55 on slim image effort, I saw this: https://dzone.com/articles/minideb-a-minimalist-debian-based-docker-image 16:35:12 I think we can grab the idea which is clean up image after install anything? 16:35:28 duonghq: impossible if you have layers 16:35:36 in general: we can have some script to wrapper thing like this 16:35:41 then you will just double cleanup size really 16:35:54 that'd be great when docker --squash is a thing 16:36:21 hmm, another idea: will we use slim image like alpine? 16:36:40 package availability becomes issue 16:36:45 I think that was brought earlier as well 16:36:53 agree with inc0 16:37:04 that being said, this creates wrapper around apt-get install it seems 16:37:11 so you install and cleanup in same layer 16:37:20 yeah 16:37:21 which is cool, but might be longer 16:37:39 but I think it's worth if we can slim down the image 16:38:14 yeah, depends on priorities I guess, but definetly worth to try 16:38:22 ya 16:38:44 duonghq: wanna try to do it? I think it should be as easy as changing macro install packages and using base distro debian 16:39:03 I think I'll give it a try 16:39:26 cool, thanks 16:39:41 cutting megs from our images is always good 16:39:49 sure 16:39:58 after docker squash works we can get even better at it 16:40:11 I'm waiting for it 16:40:27 great idea, thanks 16:40:45 another thing - with fresh images in dockerhub time of build becomes less of a factor 16:40:54 size become more 16:41:13 so if we can get slow building small images, as long as we build->upload them nightly 16:41:15 it's cool 16:41:59 so, anything else? 16:42:07 from me: no 16:42:18 anyone? 16:42:59 ok, I think we can wrap it up:) 16:43:03 thank you all for coming! 16:43:07 #endmeeting kolla